Misuse of AI is mass stupidity. The problem is that people don't understand how these things work, and the technology has been oversold (as nearly every new tech is) as something more powerful and more trustworthy than it actually is.

AI is incredibly useful. I'm already getting a ton of use out of it. But you have to treat it like an untrustworthy source, or at least have a "trust but verify" attitude. You also have to understand that it is not sentient, doesn't "care" about you, and is just a hugely powerful autocomplete engine. Any sense of intimacy or understanding you have with it is an illusion.

In engineering I treat it like a junior intern that is very fast, has memorized a huge amount of info, but makes mistakes and has to be hand-held and anything they produce must be examined and tested.

▲

bluefirebrand 6 hours ago | parent [-]

> In engineering I treat it like a junior intern that is very fast, has memorized a huge amount of info, but makes mistakes and has to be hand-held and anything they produce must be examined and tested

So you spend your own experienced dev time chasing down a high volume of junior dev mistakes instead of writing high quality experienced dev code?

Sorting through large amounts of junior intern code does not seem like a valuable use of an experienced dev's time.

▲

api 6 hours ago | parent [-]

I use it where it saves time and don't use it where it doesn't, and have learned over time where those places tend to be. As AI gets better it'll get useful in more places.

Like any dev tool you have to play with it.

▲

mwcampbell 2 hours ago | parent [-]

Are there places where it reliably saves time, though? Unlike most software tools, LLMs are stochastic, right?

	▲	bluefirebrand 2 hours ago \| parent [-]
		In my experience they reliably waste much more time than they save